In partnership with
A Doom bot powered by Jev makes about ten decisions per second, costs nearly $7 an hour, and plays worse than a conventional script. Another project sorted 1,018 research papers for $0.08. Which one looks like a real product?
This week
Jev, the classifier TypeSafe launched on September 15, does not write anything. You give it a state and closed questions, and it returns probabilities for $0.042 per million input tokens, with output free. Ten days after launch, the independent shipwithjev catalogue listed 551 projects. The demos got the attention: 66 games and real-time experiments, ten trading bots, and a Doom bot running at ten requests per second.
The uses that could make money are much less photogenic. One pipeline classified 1,018 papers into 24 themes for $0.08, compared with $3.99 for the LLM summaries in the same pipeline. One candidate matched a resume against all 6,245 Y Combinator companies in 25 seconds for $0.37, then came away with 156 founders to contact. Metaview added Jev to all of its agents over a weekend and reports candidate searches about ten times faster at the same accuracy. That is the only named production report among the 551 projects, so I would not treat those figures as an independent benchmark yet.
My verdict: use Jev when the decision is narrow, repeated at scale, and recoverable when it is wrong. Ticket triage, model routing, and a shell guardrail pass that test. Trading and automatic file deletion fail the third criterion because a bad decision is hard to undo. Typed output guarantees the form of the answer, not whether the answer is right.
Tomorrow morning, pick one task where your agent calls a large model to answer yes or no. Measure 100 decisions with your current system, then replay them through Jev on a set where you already know the correct answer. Calibrate the threshold on your own data, and send low-confidence cases to a human or the more expensive model. If you cannot define the correct answer or the cost of a mistake, do not replace anything.
Chemical-Free Farming Is Now Open To Investors
For decades, farmers have relied on herbicides and heavy equipment to control weeds.
Greenfield Robotics is working to change that.
BOTONY robots navigate crop rows and mechanically remove weeds, giving farmers an alternative to blanket herbicide applications and intensive tillage.
The potential benefits are significant:
Clean food supply
Reduce reliance on agricultural chemicals
Reduce unnecessary soil disturbance
Help farmers protect crops from weed competition
Support more sustainable farming practices
And the market is enormous.
Greenfield estimates more than 250 million acres of broadacre cropland in the U.S., with more than 100 million acres in its initial serviceable market.
With the Reg A+ now live, investors can participate in Greenfield Robotics' growth story.
This Reg A+ offering is made available through StartEngine Primary, LLC, member FINRA/SIPC. Please read the Offering Circular and related disclosures before investing. This investment is speculative, illiquid, and involves a high degree of risk, including the possible loss of your entire investment.
In the news
TypeSafe opened Jev sign-ups on September 20, then paused them on September 22 after exceeding capacity. Access remains limited, and TypeSafe says it cannot yet prove that the $0.042 per million token price is not subsidized. Any return-on-investment calculation should allow for a different price.
This week on the blog
551 Things People Built With Jev in 10 Days. The Ones That Will Make Money Are Boring.: the catalogue, the reported numbers, and the three-part filter that separates demos from usable systems.
How to Rank in Google AI Overviews Using NotebookLM (SEO Parasite Method): how a public NotebookLM page earned an AI Overview citation in under 24 hours, then got deindexed.
Revenue Stack: September 2026
This reader shows what I love about building with AI: start with a real need, break the work into small changes you can verify, and end up with an object that works outside a demo. My book Vibe Coding, For Real teaches that approach in eight steps through the build of a real application. It is not a firmware manual; it is the guide I wish I had for moving from "AI made something" to "I can actually use this." If you have a tool or device idea that only exists in your head, start by learning how to scope it, build it, and check it.
Where would you trust Jev with a decision?
Phil
The difference between the Doom bot and the 1,018-paper sorter is not speed. It is that a company already pays a person or a large model to sort those papers, while nobody pays $7 an hour to watch a bot lose to a script.
PS: which binary decision still costs too much in your system?
You're receiving this because you signed up at rentierdigital.beehiiv.com.